GKE application reports frequent errors from a component; no logging/monitoring is enabled and the issue is not reproducible. To diagnose with minimal disruption, what should you do?

Enabling Cloud Operations for GKE (formerly Stackdriver) directly on the existing cluster is the most efficient way to diagnose the issue with minimal disruption. This integrates logging and monitoring capabilities, allowing you to inspect Pod logs through the GKE Monitoring dashboard without recreating resources or migrating workloads. Creating a new cluster and migrating Pods (options B and D) introduces significant disruption and is unnecessary for enabling basic logging. Deploying Prometheus (options C and D) is an additional step for metrics collection and alerting, but the immediate need is to access logs for diagnosis, which Cloud Operations for GKE provides directly.